The Kyma system from Symbolic Sound has long been recognized as the most powerful and versatile sound-design workstation around. Combining an elegant and flexible graphic programming language and a black box full of reconfigurable DSPs, Kyma is a sampler, a synth, an effects box, a hard-disk recorder, and a live-performance environment. Version X, almost two years in the making, adds a number of enhancements to nearly every area of the system.
In this review, I'll focus on the new features in version X and the enhancements that were introduced since my review of Kyma 5.0 in May 2001. First, though, I'll give a quick overview of the system.
AROUND AND ABOUT
Kyma couples a graphical sound-programming interface (called the Kyma Language for Sound Specification) with a 3U hardware accelerator called the Capybara (see Fig. 1). The Capybara comes with 4 Motorola 56309 processors and 96 MB of RAM, and can be expanded to as many as 28 processors with a total of 672 MB of RAM. The Capybara does the processing for the system, taking the load off the host computer. Kyma runs under all recent Mac and Windows operating systems. The newly released Flame FireWire interface (included with the base system) enhances communication between the host computer and the Capybara, which allows you to play more disk tracks at the same time, among other things. OS X, Windows 2000, and Windows XP require the Flame interface, and Windows 98 SE and Mac OS 8.6 through 9.2.2 can still use the new software with a PCI slot or a PC Card.
Kyma's interface is modular and highly configurable (see Fig. 2). The areas you would use most often during a work session are the Prototype Strip (shown at the top of Fig. 2), which has access to nearly 170 basic sound-producing modules (which Kyma calls Sounds) and the Sound Browser window (on the right side of Fig. 2), where you'll find more than 1,000 preset examples (as well as your own Sounds and all supported audio files on your system). Typically you will use the multitrack Timeline (shown in the middle of Fig. 2) to arrange a sequence of Sounds or to configure Kyma to transition from one function to another.
Kyma's other important work area is the Sound Editor, which is the window you use to edit Prototype Sounds or to configure new Sounds. The Sound Editor reveals the internal structure of a Sound and shows the specific parameters it contains (see Fig. 3). A single icon can represent a network of many individual Sounds, and it's easy to expand an icon to see what it's made of. Loading Sounds into an Editor is much easier in the new version: just double-click on a Prototype, and it will automatically open in a new window. (In the past, you had to create a new Editor window and drag a Prototype into it.) In addition, you can now easily zoom in or out in the Editor. That is especially useful when you are working on a complex structure that extends past the current window.
Kyma can be used in real time under MIDI control like a traditional synth or sampler, and it can record its output to the hard drive as a WAV, AIFF, or SDII file at a variety of sampling rates and bit resolutions. The system can be configured for as many as 8-channel analog or digital output, and both S/PDIF and AES/EBU digital I/O are supported. Although you can't use Kyma's hardware with other audio applications, Symbolic Sound hopes to release drivers that will have that capability later this year.
HAVE A LOOK
The first thing you'll notice when you load Kyma X is the interface's new look. The icon for a Sound more accurately displays its purpose. For example, a Sound that accepts another signal as input has an icon that is indented on the left. A Sound with no input has a flat left edge (see Fig. 2).
Other new graphical elements offer a clearer picture of what's happening in your designs and improve functionality. For example, if you have a noise generator feeding a filter and you want to mix a sample into the filter's input as well, just drop the Sample Sound onto the “+” icon that appears to the left of the filter icon. Kyma will add a Mixer Sound that accepts both the noise generator and the sampler as input. A variety of other small enhancements — such as new icons, clearer text, and better spacing of the icons in the Sound Editor — all contribute to a more modern look and feel as well as better ergonomics.
There are only a few new Prototype Sounds in Kyma X, but when they are combined with the Sounds released since our last review, they are important additions. Analog enthusiasts will appreciate the new voltage-controlled filter (VCF) Sound, which you can use to build “legacy” synths. The VCF has amplitude, cutoff frequency, and resonance controls, and you can modulate its cutoff and amplitude at audio rate. (When's the last time you generated AM or FM with a filter?) The VCF is also characterized by a nonlinearity in its feedback loop that models the resonance of an analog filter.
Another important addition is a synthesis method called Aggregate Synthesis. Aggregate Synthesis resembles additive synthesis; however, it extends additive synthesis by summing not only sine waves, but also more complex sounds. Like additive synthesis, you can derive the control data for your aggregate sounds from a spectral analysis of a preexisting sound (see Web Clip 1). Imagine a trombone broken into hundreds of harmonics, and upon resynthesis, each harmonic controls not just a single sine-wave oscillator, but an entire grain cloud. Sound interesting? You'll find a preset included with the system. Among the other examples are Sounds that combine oscillator banks, filter banks, and formant banks. The range of sounds you'll get from the examples is broad, and like everything else in Kyma, you can tweak every Sound's parameters and customize them to your liking.
Kyma has always had the most powerful morphing capabilities I've ever seen, but the new Morph (Stereo) Prototype makes it even easier to cross the spectrum of two sounds by offering a separate oscillator bank for the left and right channels. The data for Morph is typically generated when doing a spectral analysis on a stereo file.
The old Gain and Attenuator modules have been combined into a single module called Level, and there's a new InputOutput Characteristic Sound, which serves as an arbitrary transfer function for whatever needs you might have, such as making your own compressor or nonlinear keymappings. In addition, several new modules in the spectral realm give you even more ways to process analysis data. There are also new I/O Sounds that better manage Kyma's eight audio ins and outs.
To get a sense of how the Prototypes can best be put to use, check out the included presets — there are 1,000 of them, many of which are new or enhanced. The Kyma Overview library is a comprehensive collection that makes for a good starting point. Here you'll find examples that demonstrate time-stretching, filtering, granular effects, beat munging, and almost every type of synthesis and sound-processing method you can think of. There are sequencers, vocoders, and presets that use the amplitude of an incoming audio signal to determine the playback rate of a disk file, dopplers, distortions, and morphs, among other things (see Web Clips 2, 3, and 4).
The Scripts folder is full of tools for generating lots of sound with little effort. For example, Scripts can generate a large number of MIDI events, pick randomly from a list of sample files, and convert an ASCII text file into a series of control parameters for anything you want. In some cases, Scripts have variables that you supply when you first run a Sound. These could be the number of filters you wish to use in a process or the length of time that you want a process to run. The variables allow you to create endless variations on the results of the processes that the script is running.
One of the Scripts presets, called One Sound File Follows Another, is a clever playlist-style example that prompts you for a number of files when it runs, and then opens a file-browser window for you to pick the number of files you've designated. It then plays all of the files in succession. Another preset, called Random Sound Builder, will combine any number of source Sounds that you give it in random permutations (see Web Clips 5 and 6).
Expressions (found in the Expression Library folder) are somewhat like Scripts in that they can provide control data for any parameter you choose, but they're also useful for processing existing controls. In some ways similar to the Functions feature of the Kurzweil K2000-series instruments, Expressions include various types of random-number generators, arrays, tools to manipulate tempo or beats per minute, and other types of function generators. The big news here is that you no longer have to type “Expressions” into a parameter field, because you can drag them directly from their folder to a field.
Expressions offer great potential for experimentation. For example, I made several Sounds that used the new nextChaotic Expression to play back random segments from a sample. The first example chose a new segment on every beat, and the second picked a new segment every time a MIDI Note On was received. In both cases, the length of the segment was under the control of a parameter that could be updated in real time.
Kyma's Replaceable Input feature is not new, yet it is still worth a mention. The Overview Library includes hundreds of effects, and if you want to hear how one of your samples will sound when processed by the effects, simply designate your sample as the substitute input by selecting it in the Browser. Then, as you go through the effects, they will all use your sound in place of the sound they process by default. You can also designate Kyma's live audio input as the substitute sound.
Kyma's developers have included libraries from earlier versions, giving you even more ready-made patches to play with. Fortunately, the lion growl morphing into a Harley-Davidson engine sound — one of my all-time favorites — is still around (see Web Clip 7).
Kyma's Timeline has been a central feature of the program for several years, and a number of enhancements make it an even more powerful environment for building your compositions (see Fig. 4). For starters, you can now render only a selected range across selected tracks, which makes previewing sections of longer works much more efficient. Of course you can still solo or mute entire tracks.
The Disc Caching feature is now built into the Timeline, which makes it even easier to access. Disc Caching prerenders a section (typically a complex one) to disc and automatically plays it as an audio file. That significantly lightens the real-time load on the processors and can be useful when you're using Kyma in live performance.
Navigation in the Timeline has also been improved. You can now drag markers around more easily, and you can associate Timeline markers with MIDI Program Change messages. This gives you nearly unlimited ways to move around within a long composition while you are composing or performing in a live setting.
Kyma includes a number of prebuilt Timelines, which are, in many cases, large-scale works for which you can substitute your own source Sounds or samples. Some of the examples feature synthesis processes that crossfade into others over time, and you can easily change the values of any of the default parameters to customize the processes. Like Iannis Xenakis's UPIC system, you can build a Timeline that sequences a large number of events, and then alter the time it takes to run the entire set of processes.
One of the keys to real-time control in Kyma is the versatile Virtual Control Surface (VCS), a panel that pops up any time you play a Sound that has a parameter that can be adjusted while the Sound plays (see Fig. 5). New enhancements to the VCS include the ability to copy and paste controls from one VCS to another and manipulate the size of the VCS window more easily. New keyboard shortcuts are included for several VCS features as well as some tweaks that make using the VCS Editor easier. Like the Timeline, VCS presets can now be selected with Program Change messages.
Buttons and knobs can be color coded for easier identification. If you have two Sounds on two different MIDI channels, both containing a control with the same name, color coding them will help you to differentiate between the two.
Kyma's Tool menu includes some of the most interesting features in the program. Here you'll find functions to build microtuning tables, “bundle” a Sound and all the sample files it might contain for saving or moving to another computer, record external audio to disk (the 24-bit converters sound excellent), and perform a variety of analysis functions.
One of the new Tools, called mouseToEventValues, allows you to control two parameters by sweeping your mouse around an x-y grid. Although that is fairly commonplace in other synthesis environments (Reaktor comes to mind), this feature gives Kyma a much-needed enhancement in the realm of graphic real-time control, an area that I hope will continue to evolve. Although you can't record your mouse movements directly into the mouseToEventValues interface, you can record them using the Timeline.
Current users might not be aware that you can create graphical envelopes for any parameter by dragging the Graphical Envelope Sound and dropping it directly into the parameter field of your choice. Graphical Envelopes can have numerous segments, and their scale and rate values can be altered in real time.
Another new tool, called Logistic Map, produces repeating rhythmic patterns that can be sent into total (or controlled) chaos with the push of a slider. It generates a real-time animated visualization of the chaotic function, leading me to wonder if graphics will become a larger part of Kyma's repertory in the future (one can only hope).
Kyma remains the most powerful analysis/resynthesis environment around. Where else can you do a real-time spectral analysis on an incoming audio signal and perform a cross-synthesis of the live sound with a sound on your hard drive while tracking the amplitude of the live sound and using it to scale the level of the disc file? Try doing that with your favorite VST plug-in.
Version X enhances Kyma's analysis capabilities by giving the Spectral Analysis Tool the ability to automatically guess the fundamental frequency in any sound you use as input. This can make the analysis routine more accurate than the trial-and-error method formerly required. You can also analyze stereo files, and Kyma now uses more oscillators in the resynthesis stage than in the past. (You can reduce the number of oscillators being used if you find that the resynthesis is taxing your system.)
An extensive tutorial in the documentation tells you all you need to know about creating your own Tools. Custom Tools give you a great way to build your own unique interfaces to existing Sounds, design interactive installations, and create recording or file-playback control surfaces.
MODEL THE PHYSICAL
Although it has all of the requisite tools for building physical models, Kyma doesn't include many examples that use this synthesis method. However, a large number of physical-modeling Sounds submitted by Kyma users have recently appeared on Kyma's Web forum (in the Community section at www.symbolicsound.com).
Among the best physical models are David McClain's plate and membrane simulations, which use various combinations of delays, attenuators, noise generators, and oscillators to create effective real-world simulations. By tweaking the parameters (and using Kyma's Roll the Dice parameter randomizer), I created a range of sounds from sharp metallic and delicate twinklings to struck bamboo (see Web Clip 8). Other user-contributed models are examples of plucked strings and analog synths. A massively powerful gem called the Articulator produces a vast range of vocal timbres. Contributed by an anonymous user, the Articulator is not pure physical modeling (it uses a sample at its core), but it imposes a convincing vocal-like timbre on whatever sample you use as a source. Other examples include an interface for linking Kyma with Cycling '74's Max/MSP and filter-building toolkits.
WHAT'S UP, DOC?
Kyma has always had extensive printed documentation, but the previous manual was a bit advanced for the average user. The printed manual for Kyma X is more appropriate for those just learning the system, walking the user through a vast number of areas and topics. The 400-page tome covers every major aspect of the program and has extensive reference materials. Though it may take you hours to get through the examples, it is well worth the effort.
Of course you don't need to be a computing guru or an accomplished electronic-music composer to take advantage of what Kyma has to offer. In fact, it takes only a short amount of time to go from installing the system to producing good results. With just a few mouse moves, you can configure a multilayered Timeline that lasts 2 (or even 20) minutes. Playing and tweaking the preset examples will keep you busy for hours.
THIS YEAR'S MODEL
Kyma remains the most mature, robust, and powerful sound-design workstation anywhere. For its price — about the same as a fully loaded hardware sampler — it is an unbelievable deal.
Although the world has mostly moved to the desktop for much of its synthesis, sampling, and recording needs, Kyma X is a perfect example of how hardware-accelerated tools can still beat software by a wide margin. There is no software or bundle of soft tools that can come close to the power, expandability, and flexibility that Kyma offers. If you haven't heard the demos at the Symbolic Sound Web site (www.symbolicsound.com/hearkyma.html), give them a listen to get a taste of the system's abilities.
Associate EditorDennis Millerrediscovered the joys of Kyma while doing this review and is now uninstalling all of his other software.
Kyma X sound-design workstation basic workstation $3,570 audio I/O upgrade $995 expansion card $595
FEATURES5.0EASE OF USE4.0QUALITY OF SOUNDS5.0VALUE5.0RATING PRODUCTS FROM 1 TO 5
PROS: Huge library of examples. Nearly 170 basic prototype modules. Extensive real-time control. Excellent printed manual and tutorials.
CONS: Hardware not accessible by other audio software.
Kyma X Specifications
Analog Outputs(4) XLR (expandable to 8)Digital I/O(2) XLR AES/EBU or (2) RCA S/PDIF (RCA)Other ConnectionsMIDI In/Out/Thru; (1) BNC word-clock input;
(1) BNC house sync; (1) RCA VITC In/Out;
(1) RCA LTC In/OutOutput Level+14.5 dBuInput Clipping Level+14.0 dBuResolution24-bitInternal Processing24-, 48-, 56-bit (algorithm dependent)Sampling Ratesall standard rates between 32 and 100 kHzDynamic RangeA/D: 110 dB (unweighted); D/A: 107 dB (unweighted)Frequency Response20 Hz-20 kHz (+0.04/-0.26 dBu @ 44.1 kHz)Input Impedance10 kžCrosstalk-110 dBNoiseA/D: 110 dB; D/A: 105 dBTuning Resolution0.0026 HzPrototype Algorithms170 unique basic modulesFactory Patches> 1,000Dimensions16.5" (W) × 5.5" (H) × 17" (D)Weight15 lbs.